canvascoding/canvas-notebook / seed_skills/docx
docx
UnverifiedUse this Canvas-owned clean-room skill for creating, reading, editing, reviewing, redlining, commenting, or repairing Word `.docx` documents. Trigger for Word documents, DOCX files, memos, reports, letters, templates, tracked changes, comments, table of contents, headers, footers, page numbers, and Google Docs-targeted documents that should be authored as DOCX first.
DOCX Documents
This skill covers Word-compatible .docx work in Canvas Notebook. It is a
clean-room Canvas skill; do not copy vendor skill material, hidden prompts, or
proprietary helper code into this folder.
Core Contract
- Always edit a checked-out working copy. Call
checkout_docxwith the workspace-relative input path, edit only its returnedworkingPath, and publish withcommit_docx. This applies even when overwriting is requested: the commit tool checks the original revision and edit lease. - For a new document, use
checkout_docxwithcreateOnly: trueand the desired.docxpath. Create the returned working file with Python, then commit its checkout ID. An existing output is never silently overwritten. - Use
CANVAS_AGENT_TEMP_DIRfor scripts, working copies, unpacked OOXML, conversions, and render previews. Publish DOCX working copies throughcommit_docx; usecopy_pathormove_pathfor requested non-DOCX exports. - Preserve existing document style for targeted edits. For new documents, set explicit page size, margins, paragraph spacing, font choices, headings, tables, headers, and footers instead of relying on Word defaults.
- Prefer editable Word-native content: paragraphs, runs, tables, lists, sections, headers, footers, comments, and fields. Do not turn pages into screenshots unless the user asks for an image-only deliverable.
- Keep intermediate render images, PDFs, unpacked XML, and scratch files out of the final response unless the user asks for them.
- Store final user-facing files in the requested workspace path using a stable descriptive filename.
Recommended Tools
python-docxfor normal document creation and structured edits.- Standard library
zipfilepluslxmlfor OOXML edits thatpython-docxcannot express, such as true comments, content controls, fields, or targeted XML repair. - Pandoc for simple, text-oriented Markdown-to-DOCX or DOCX-to-Markdown conversion. Do not use it for format-faithful edits to an existing DOCX.
- LibreOffice
sofficefor DOCX-to-PDF conversion. Before invoking it, read references/headless-libreoffice.md. - Poppler
pdftoppmorpdftocairowhen available for PDF to PNG rendering.
Use the installed runtime dependencies. Agent shell subprocesses can write only to the current session temporary directory; workspace files are read-only there. Do not try shell redirects, direct Python saves, copies, or custom scripts to bypass publication. Never store API keys or secrets in a skill directory.
Checkout, publication, and recovery
- Keep the
checkoutId,workingPath, andlockExpiresAtreturned by checkout. Useinspect_docx_checkoutwithrenewLease: truebefore the lease expires during longer edits. Expired or replaced leases are never recreated by renewal. - Finish and await all Python/render subprocesses before
commit_docxreads the working copy. Pass only the checkout ID; never supply or replace its baseline. - A committed result is idempotent. If publication returned a transient error
with status
prepared, retry the same checkout ID: its exact proposed bytes remain fixed even if scratch files changed afterward. - If status is
conflict, report the conflict and itsrecoveryId. The proposal is retained outside temporary cleanup. Useinspect_docx_checkoutwithrestoreWorkingCopy: trueto recover it in the same task; this replaces only the scratch copy. To keep a separate proposal, create a newcreateOnlycheckout and copy the recovered content into that checkout's working path. - Use
release_docx_checkoutwhen stopping without publication. It retains the current draft and releases the lease. Released checkouts cannot commit. - If the lease belongs to another editor, explain who is editing when that information is available and preserve any proposed work. Never force release another editor or overwrite a newer version to complete a document task.
Workflow
- Identify whether the task is read/review, create, targeted edit, structural repair, comments/redlines, or conversion.
- Check out the input or reserve a new output. If editing an existing document, inspect the working copy's current structure before changing it: headings, sections, tables, images, comments, tracked changes, headers, footers, and metadata.
- Make the smallest reliable edit that satisfies the request. Use styles and reusable helper functions instead of scattered one-off formatting.
- For visual deliverables, render the DOCX when possible and inspect page images for clipping, overlap, missing glyphs, broken tables, bad page breaks, header/footer issues, and inconsistent spacing.
- If render output is wrong, fix the DOCX and render again. If rendering is unavailable, perform structural checks and clearly state that visual QA could not be completed.
- Publish the validated working copy with
commit_docxand verify that the returned status iscommittedbefore reporting the workspace file as saved.
Common Tasks
- Read or summarize: extract text with
python-docx; inspect tables and images separately; render if layout matters. - Create a new document: build with
python-docx, define styles up front, then add content. Use real tables and lists. - Targeted edit: preserve existing formatting. Avoid document-wide restyles unless asked.
- Comments: use OOXML comments only when true Word comments are required. Otherwise, a review memo or inline markup may be safer.
- Tracked changes: true Word redlines require OOXML patching. If exact revision semantics matter, validate in Word-compatible tooling after writing.
- Google Docs target: create a clean
.docxfirst. Avoid Word-only visual effects that import poorly, such as decorative title rules, complex floating objects, and brittle fields.
Final Checks
- Final DOCX opens without repair warnings.
- No placeholder tokens, internal notes, or unresolved TODOs remain.
- Tables fit within margins and have readable column widths.
- Headers, footers, page numbering, links, and references are intentional.
- For high-stakes documents, mention any environment limitation that prevented render or Word-native validation.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| canvas-designanthropics/skillsCreate beautiful visual art in .png and .pdf documents using design philosophy. You should use this skill when the user asks to create a poster, piece of art, design, or other static piece. Create original visual designs, never copying existing artists' work to avoid copyright violations. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| docxanthropics/skillsUse this skill whenever the user wants to create, read, edit, or manipulate Word documents (.docx files) or Word templates (.dotx files). Triggers include: any mention of 'Word doc', 'word document', '.docx', '.dotx', or requests to produce professional documents with formatting like tables of contents, headings, page numbers, or letterheads. Also use when extracting or reorganizing content from .docx or .dotx files, inserting or replacing images in documents, performing find-and-replace in Word files, working with tracked changes or comments, or converting content into a polished Word document. If the user asks for a 'report', 'memo', 'letter', 'template', or similar deliverable as a Word or .docx file, use this skill. Do NOT use for PDFs, spreadsheets, Google Docs, or general coding tasks unrelated to document generation. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| pdfanthropics/skillsUse this skill whenever the user wants to do anything with PDF files. This includes reading or extracting text/tables from PDFs, combining or merging multiple PDFs into one, splitting PDFs apart, rotating pages, adding watermarks, creating new PDFs, filling PDF forms, encrypting/decrypting PDFs, extracting images, and OCR on scanned PDFs to make them searchable. If the user mentions a .pdf file or asks to produce one, use this skill. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| pptxanthropics/skillsUse this skill any time a .pptx or .potx file is involved in any way — as input, output, or both. This includes: creating slide decks, pitch decks, or presentations; reading, parsing, or extracting text from any .pptx or .potx file (even if the extracted content will be used elsewhere, like in an email or summary); editing, modifying, or updating existing presentations; combining or splitting slide files; working with templates (.potx), layouts, speaker notes, or comments. Trigger whenever the user mentions "deck," "slides," "presentation," or references a .pptx or .potx filename, regardless of what they plan to do with the content afterward. If a .pptx or .potx file needs to be opened, created, or touched, use this skill. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |